Missing Files for create_tfrecord.py

[vsltech]

Please share how to create_tfrecord & supply input image fro demo.py? Also, these files are missing '/share/zhui/mnt/demo_20.tfrecord' from input_data.py

[goodhamgupta]

Hi @vsltech,

Unfortunately, it seems like the demo.py will work for images that have numbers between 0-10. My best guess would be to create the tfrecord for the MNIST dataset and try running the script.

The synthetic characters dataset is available here. Once the dataset is downloaded, extract the dataset and do the following:

• Create a train.txt file with lines containing the path of the images you want to train on and the groundtruth_text. For example,

./mnt/ramdisk/max/2998/7/100_KINDERGARTNERS_42516.jpg KINDERGARTNERS
./mnt/ramdisk/max/2998/7/101_MILLENNIAL_48450.jpg MILLENNIAL
./mnt/ramdisk/max/2998/7/102_multiplicities_50355.jpg multiplicities

• Run the create_tfrecord.py script with the following arguments:

  python3 create_tfrecord.py --output_path path_of_tfrecord_file --tags_file_path path_of_train_txt_file

• Once the tfrecord file has been created, run the train.py file with the following arguments:

  python3 train.py --tfrecord_file_path path_to_tfrecord_file

Let me know if you have any issues! Thanks!